Large-scale context for the UK floods in summer 2007

Transcription

1 Large-scale context for the UK floods in summer Weather September, Vol., No. Mike Blackburn John Methven Nigel Roberts National Centre for Atmospheric Science, University of Reading Department of Meteorology, University of Reading Met Office, Joint Centre for Mesoscale Meteorology This article shows how the rainfall distribution over the UK, in the three major events on June, June and July, was related to troughs in the upper-level flow, and investigates the relationship of these features to a persistent large-scale flow pattern which extended around the Northern Hemisphere and its possible origins. Remote influences can be mediated by the propagation of large-scale atmospheric waves across the Northern Hemisphere and also by the origins of the air masses that are wrapped into the developing weather systems delivering rain to the UK. These dynamical influences are examined using analyses and forecasts produced by a range of atmospheric models. Other papers in this Special Issue, Hanna et al. () and Prior and Beswick (), examine observations of the extreme rainfall events in greater detail. Spatial and temporal extent of the rainfall anomaly Heavy rainfall extended throughout much of June and July and was not confined to the UK. Figure shows precipitation anomalies as a percentage of the climatological average for the months of June and July. The high rainfall totals extended across much of western and northern Europe, excluding Scandinavia. In contrast, Mediterranean Europe experienced exceptionally dry conditions, particularly in July and running into August (not shown), which were associated with higher-than-average temperatures and severe fires in Greece. The regional extent of the precipitation anomalies suggests that large-scale atmospheric processes played a role. The similarity of the anomaly distributions for June and July also suggests that the synoptic regime was rather persistent over these two months. Figure illustrates the location and movement of lows and highs in the mid-latitude upper troposphere using a time-longitude diagram. Looking down the time axis at the Greenwich meridian, two distinct periods are evident: Figure. European precipitation anomalies during as a percentage of the climatological average. (a) June, (b) July. (Courtesy, National Oceanic and Atmospheric Administration, not subject to copyright protection.)

2 Figure. Time versus longitude diagram showing the evolution from May to August of geopotential height anomalies, in metres, averaged over N near the tropopause ( hpa). Dates of the three UK rainfall events are labelled A C along the Greenwich meridian. Data are from NCEP operational analyses. June and July, when a trough (low geopotential height anomaly) was almost stationary over the UK. Three extreme rainfall events were embedded within these two periods, producing severe flooding in Northern England and the West Midlands on June and June and in southern and western England on July. In between these two stationary periods there was a mobile period ( July) when several synoptic scale systems propagated across the UK. Wave pattern on the mid-latitude jet stream The persistent upper-level trough over the UK was associated with an unusually persistent Rossby wave pattern on the mid-latitude jet stream. Figure (a) shows wind at hpa averaged over the period June to July, which corresponds to the period of troughs over the Greenwich meridian in Figure. Even for this long averaging period, the trough over the UK and Ireland was clearly associated with cyclonic (anticlockwise) flow diverting the jet stream equatorwards over northern Spain and southern France. In the climatological average between these dates, the jet stream is split in the European sector (Figure (b)), with a jet exit over Ireland and the UK and a subtropical jet entrance over North Africa. Typically in summer, midlatitude weather systems track north of the UK towards Norway, crossing the jet exit towards the north-east. However, in summer, the weather systems tracked slowly into the southern UK and France and the heaviest rainfall, as we shall show, was associated with warm, moist air from the south wrapping around the eastern flank of these cyclones over the UK. Perhaps even more remarkable is that the wave pattern was almost stationary around the entire Northern Hemisphere (Figure (a)). Looking upstream from Europe, the jet stream was diverted equatorwards around distinct troughs over eastern North America, the Gulf of Alaska, the west Pacific, north-west China and the Middle East, projecting strongly onto Zonal Wavenumber. This pattern was much stronger during the stationary trough periods (not shown). Unusual weather events were connected with other quasi-stationary features in the wave pattern. For example, in Iceland, June was warm and an unusual dry spell began in the south and west on the th with Weather September, Vol., No. Figure. The Northern Hemisphere jet stream, shown by the strength and direction of the wind at hpa in ms -, time-averaged over the period June to July, (a) for and (b) the climatological average for.

3 Weather September, Vol., No. only a few millimetres of precipitation falling until the end of July (Trausti Jonsson, Icelandic Met. Office). Iceland lay beneath the ridge upstream of the trough over the UK. There was heavy rainfall in Texas during both late June and late July, associated with southwesterly flow ahead of the trough over eastern North America. The eastern Mediterranean heatwave beneath a ridge has already been noted and the western United States also experienced heatwave conditions during early July, downstream of the trough positioned west of the coast there. Since the UK rainfall events were associated with persistence of an upper-level trough, it is natural to ask why the hemispheric Rossby wave pattern recurred. Once such a Rossby wave has been generated it propagates westwards relative to the eastward flow of the jet stream and can become stationary when its propagation rate is equal and opposite to the flow (averaged in some way across the jet in latitude and height). Therefore, it is possible for such a wave pattern to exist in the absence of any forcing (a free mode). However, stationary Rossby waves can be generated by a number of processes: by large-scale divergence in the upper troposphere associated with condensation and latent heating (Sardeshmukh and Hoskins, ) and by mountains and temperature contrasts along the ground (e.g., land/sea contrasts) (Hoskins and Karoly, ). If a wave source switches on, the stationary wave pattern will grow, with wave energy radiating away from the source at the group velocity. Where the atmospheric flow is weak, Rossby wave energy propagates along great circles, although the waves can be refracted and bent along the mid-latitude jet stream, which acts as a Rossby wave guide (Hoskins and Ambrizzi, ). The fact that the mid-latitude jet stream was not broken in the European sector but connected around the Northern Hemisphere would have been conducive to trapping Rossby wave energy. The connectedness of the jet across Spain, however, was associated with the existence of the wave trough there, so the above argument is consistent but not causal. Global events What might have influenced the midlatitude Rossby wave pattern? In Figure (a) it is illustrated in terms of stream function anomalies (stream function resembles geopotential height but gives greater emphasis to the Tropics). The most obvious features are a train of waves appearing to arc across the Pacific from the Philippines to North America and a dipole over the east Atlantic. Similar wave trains over the Pacific and North America have been shown to be associated with tropical heating anomalies acting as a Rossby wave source the dominant pattern of variability is the Pacific Figure. Anomalies from climatology averaged from June to July. (a) streamfunction near the tropopause in ms- (sigma=.); (b) outgoing longwave radiation (OLR), in Wm-, from interpolated NOAA data; (c) sea surface temperature anomaly in Celcius. North American (PNA) pattern (Wallace and Gutzler, ), but this emanates from the tropical mid-pacific. In addition, it is not clear that the Atlantic dipole would be related. The amplitude of the wave pattern was largest across North America (Figure (a)), with a ridge over the Rockies, which is consistent with forcing of Rossby waves by the mountains there. Animations reveal that the Rossby wave pattern generally propagated

4 (a) (b) Weather September, Vol., No. (c) (d) (e) (f) The wave pattern during the three UK rainfall events The Rossby wave pattern is most apparent on the mid-latitude tropopause a surface defined by the rapid increase in static stability from the troposphere to the stratosphere above. On average, the tropopause slopes down from the Tropics towards the Poles. Where it is lower it is associated with lower potential temperature (θ), so θ on the tropopause generally decreases towards the Poles. Since large-scale motions are almost adiabatic and frictionless, air is constrained to move along the tropopause conserving its potential temperature. Therefore contours of θ on the tropopause can be used to track the air motion. eastwards together with its attendant synoptic scale systems, but appeared to stall with the ridge positioned over western North America during the two stationary phases. It is plausible that a slowly propagating wave resonated with a wave forced by the mountains and therefore phase-locked there. This is speculative; carefully constructed numerical model experiments would be required to establish any such mechanism. Circumglobal teleconnection patterns have been observed previously in boreal summer by Ding and Wang (), who argue that they are related to variations in tropical heating, primarily in the Indian Summer Monsoon (ISM), and to the El Niño Southern Oscillation (ENSO) via its impact on the ISM. Although the positions of Ding and Wang s Atlantic and European anomalies vary through the summer season, their July pattern is similar to the hemispheric anomaly pattern in summer (Figure (a)). The La Niña phase of ENSO was beginning during summer, evident as a cold sea surface temperature (SST) anomaly in the equatorial east Pacific (Figure (c)) and a warm outgoing longwave radiation (OLR) anomaly from the atmosphere above (Figure (b)), associated with a reduction in cloud amount or cloud-top altitude there. However, the SST signal was relatively weak at this time. Greater anomalies are evident in Figure in the tropical Indian Ocean, associated with warm SSTs and cold OLR anomalies above, indicating enhanced convection in the Asian summer monsoon region. Rodwell and Hoskins () have identified a mechanism by which the warm anticyclone generated by monsoon heating interacts with the midlatitude jet stream to force descent of air from the mid-troposphere over the eastern Mediterranean. The resulting drying and reduced cloud can enhance atmospheric cooling by radiation, increasing the descent in a local positive feedback. Therefore, the dry anomalies across the Mediterranean shown in Figure could, in part, be related to enhanced convection in the Asian summer monsoon region. Figure. Potential temperature at the tropopause showing the Northern Hemisphere wave pattern for the three rain events. (a) (b) and June; (c) (d) and June; (e) (f) and July; all at UTC. The left panels are two days prior to each event and the right panels are for the day of the event. Troughs in the wave pattern are numbered with trough- over the UK. Data are from ECMWF operational analyses. If air moves equatorwards along the tropopause, it carries low θ with it. The air mass is then cold relative to its surroundings and this is associated with a cyclonic flow anomaly around it. This flow brings more cold air from the north on its western flank, while at the same time bringing warm air from the south on its eastern flank. The net effect is for the cold anomaly to propagate westwards (against the eastwards jet). This is the fundamental mechanism of Rossby wave propagation. The right column of Figure shows θ on the tropopause during the three UK rainfall events. In all cases, trough- (cyclonic flow anomaly) over the UK and France is

5 - Weather September, Vol., No. Figure. Forecasts for Case A on June, from the Met Office operational North Atlantic and European model, initialized at UTC. Mean sea level pressure (black contours in hpa); selected warm values of wet bulb potential temperature (θ w) at hpa (colour shading); and potential vorticity on the K isentropic surface (blue contours, in PV-units). The PVU contour outlines the upper-level trough. Bold dots indicate continuous rainfall exceeding mm hr - (considered heavy on the km model grid) and triangles indicate heavy showers. outlined by cold air (low θ). Similarly, the other troughs (labelled moving upstream) are readily identified as cold anomalies or extensions of air equatorwards from the high latitude polar vortex. Ridges (anticyclonic flow anomalies) are linked to warm anomalies where air moves polewards from the subtropics. The left column of Figure shows maps of θ on the tropopause two days preceding each rainfall event. Three features are striking:. The dominant troughs barely moved eastwards over the two preceding days.. The trough over the UK wrapped cyclonically (anticlockwise) in each case.. The precursor wave patterns were remarkably similar between cases, especially June and July. The three events therefore occurred in similar hemispheric settings. The following sections zoom in to the regional scale, to investigate the relationship between the large-scale trough and the rainfall events. Case A: June The upper-level trough seen in Figures (a) and (b), characterized by a cyclonic cut-off at tropopause level, moved very slowly over the UK from the south-west during this period. The cyclonic circulation extended through the troposphere, bringing warm, moist air from the south in a warm conveyor belt (WCB) across Northern England. The WCB ascended in an occluded front which wrapped around the northern side of the cyclone from the east and remained almost stationary, bringing continuous rainfall. Surface synoptic charts are shown in Figure of Hanna et al. () in this issue (p. ). The situation on the morning of June is depicted in Figure by the Met Office operational forecast, using their km resolution North Atlantic and European (NAE) model initialized from UTC. The upper-level trough, containing high values of potential vorticity (PV) and outlined by the PVU contour on the K potential temperature surface in Figure, rotated cyclonically (anticlockwise) through the period. The WCB air being drawn in from the south-east below the trough is identified by high values of wet bulb potential temperature (θ w) at hpa (c) (a)... PV (PVU) T=-. days hpa Release on UT // Wet bulb PT (C) T=-. days UT // (.N,-.E) Distance (km) (.N,.E) WCB... (colour shading). Heavy rainfall in the forecast (dots) occurred mainly on the northern side of the WCB, while heavy showers (triangles) progressed across the Midlands directly below the upper-level trough. Some of the heaviest observed rainfall rates were related to convection in this region. We next relate this limited area forecast to the global analyses and consider the origin of the warm moist air mass. Figure shows T=-. days hpa Release on UT // E.N p=hpa T=-. days UT // Longitude (degrees) Figure. Reverse domain-filling trajectory analysis for Case A, showing properties of air arriving on a dense grid of points over the depicted domain at UTC on June. (a) Potential vorticity carried along the trajectories from their origin two days previously, for trajectories arriving on the hpa level; (b) Change in pressure along two-day trajectories arriving on the hpa level, with green and blue indicating ascent and orange descent; (c) Vertical cross-section along the dashed line in (a) (b), showing wet bulb potential temperature carried along the trajectories from their origin two days previously. Warm conveyor belt flow is denoted WCB; (d) -day back-trajectories arriving in the vicinity of hpa over north-east Yorkshire, shown by an X in panels (a) (c). Data are from ECMWF operational analyses. (d) (b)

6 Box. Reverse domain-filling trajectory analysis One way to examine weather-system structure using atmospheric analyses is to calculate trajectories following many air masses backwards in time from a fine-scale D grid covering the region of interest a technique called reverse domain-filling trajectories for a D domain (RDFD, Methven et al., ). This technique reveals the detailed structure of contrasting air masses and indicates their origins. This is achieved by carrying values of conserved quantities along the trajectories from several days previously and plotting them on the dense arrival grid. For example, wet bulb potential temperature (θ w) reveals temperature and moisture contrasts in fronts, while potential vorticity (PV) reveals the shape of the tropopause. an RDFD calculation (Box ) using operational analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF), initiating back-trajectories at UTC on June from a dense grid spanning the troposphere over the UK, Ireland and north-west France. In Figure (a), points on the grid at hpa are coloured by the PV carried along trajectories from their origin two days previously. High PV values (red) indicate air of stratospheric origin. A high PV anomaly also induces cyclonic flow at its own level, but also remotely above, below and around it. The shape of the stratospheric intrusion (red) matches the shape of the PV contours in Figure (c), showing that the fine-scale structure of the weather system is determined by the stirring of the large-scale winds resolved by the analyses. Therefore we can be confident that the analysed trajectories give a good representation of the origin of air masses wrapped into the weather system. Figure (b) shows the change in pressure along two-day trajectories arriving on the grid at the hpa level negative values indicate ascent from levels with higher pressure. Strongest ascent (blue) identifies the WCB of this cyclone being wrapped around its northern side. Figure (c) is a vertical section from south to north across the cyclone (following the dashed line in Figure (b)) and shows θ w carried from the origin of two-day trajectories to their arrival points on the grid. The X symbol at hpa corresponds with the location of the X over north-east Yorkshire in Figure (b). This point is within a warm, moist air mass that has ascended strongly (note that θ w is conserved following saturated or unsaturated air motion). Figure (d) shows fourday back-trajectories from the vicinity of the X. Clearly, this air mass originated from the mid-atlantic near the Azores where it would have gained its characteristic temperature and moisture, influenced by fluxes from the ocean, giving θ w values near C, as in Figure. The surface front is also obvious km along the section (Figure (c)) running beneath the southern flank of the WCB in Figure (b). Clearly the low-level cold air to the north of the front was statically stable. However, the trajectory analysis shows that, above the front, advection tended to create a strong negative gradient in θ w with height. This indicates potential convective instability which, coupled with the largescale ascent within the WCB, would result in persistent rainfall, as seen in Figure, with heavier bursts of rain from embedded convective clouds. Case B: June This case is similar in many respects to the preceding Case A. A large-scale upper-level trough remained stationary over the UK for several days (Figures (c) and (d)), creating convectively unstable conditions which led to localized heavy rain from June (Grahame and Davies,, this issue). On the rd, a region of particularly cold air on the tropopause advected southwards from Greenland within the broader trough and wrapped up cyclonically as it reached the UK on the th. During the same period, a weak weather system moved across the Atlantic towards north-west France, evident at the surface (Weather log, June ) and as a region of relatively cold tropopause air north-west of Spain in Figure (c). The resulting cyclone intensified over the UK on the th and the NAE model shows that the surface low was directly beneath the cyclonically wrapping tropopause anomaly (Figure ). As in Case A, the heaviest rain fell along the northern flank of the upper-level trough within a belt of warm moist air (the WCB) that was wrapping round from the east. In this case the WCB ascent and resulting band of rain became almost stationary over Yorkshire, where it rained for most of the day. The ascent was associated with the coupling between the upper-level trough and the lower-level thermal gradient. This type of slow-moving synoptic situation is recognized as being conducive to high rainfall (Hand et al., ), but is more effective in summer when the air has a higher water content. The RDFD analysis shows the cyclone structure for this case to be similar to Case A, with the trough at tropopause level (Figure (a)) wrapping cyclonically, centred roughly over the south coast of England and the ascending WCB air mass wrapping into the system from the east across northern England. Back-trajectories from the WCB above north Yorkshire (the X in Figures (b) and (c), shown in Figure (d)), originated from the Azores region, as in Case A, and were characterized by a very similar moisture content and θ w. Comparison with the surface charts suggests that this WCB air was carried from the Azores in the Weather September, Vol., No. Figure. Forecasts for Case B on June, from the Met Office operational North Atlantic and European model, initialized at UTC. Details as in Figure.

7 Weather September, Vol., No. (c) (a) PV (PVU) T=-. days hpa Release on UT // Figure. Reverse domain-filling trajectory analysis for Case B, showing properties of air at UTC on June. As in Figure except (a) potential vorticity carried along -day trajectories arriving on the hpa level; (b) change in pressure for -day trajectories arriving on the hpa level; (c) the cloud head is additionally denoted CH; (d) -day back-trajectories from hpa over Yorkshire, shown by an X in panels (a) (c). warm sector of the Atlantic weather system approaching from the south-west, generating the low-level temperature contrast that coupled with the upper-level trough over the UK to produce the intensifying cyclone on the th. Note that the air-mass over the English Channel labelled CH in Figure (b), that had experienced ascent and was being wrapped back eastwards, can be identified with the cloud head and was cold air characterized by low θ w in the vertical section, Figure (c). It was therefore distinct from the warm, moist ascending air in the WCB. This analysis is corroborated by the distinctive cloud patterns in the satellite image shown in Figure of Hanna et al. () in this issue (p. ) T=-. days hpa Release on UT // Case C: July A cut-off cyclonic anomaly on the tropopause had been almost stationary over the UK for several days preceding July (Figure ), with a weakening surface low beneath it (Weather log, July ). The upper-level precursor was remarkably similar to that on June, both the hemispheric Rossby wave pattern (Figure ) and the regional upperlevel trough pattern to be shown by the RDFD analysis in Figure (a). However, the weather system at low levels did differ. Figure shows a sequence of threehour NAE forecasts from, and UTC on July. In the evening of July, very warm, moist air lay to the east of the Greenwich meridian across France and E.N p=hpa T=-. days UT // (.N,-.E) Distance (km) (.N,-.E) Longitude (degrees) Wet bulb PT (C) T=-. days UT // CH WCB (d) (b) WCB CH Benelux. The flow had a strong southerly component across the region, curving polewards around the eastern flank of the largescale trough centred south of Ireland. There was no evidence of a synoptic scale wave on the lower-level temperature front at this time. However, at upper levels, the high PV within the trough was elongated and its southern end rotated rapidly, moving from northern Spain at UTC to north-east France by UTC (Figure ). This resulted in rapid baroclinic development of a temperature wave at low levels, spinning up a weak surface cyclone with a warm sector extending from France at UTC across southern England to the Welsh border by UTC. Initially ( UTC), the heaviest rainfall in the forecasts occurred within and at the tip of the warm sector. The radiosonde ascent from Herstmonceux, south-east England, at UTC (Figure ) shows that the troposphere was saturated through its depth as the warm sector first moved over England. The column precipitable water content in this profile was almost mm, which is high compared to the climatological average over the UK of mm. This is likely to have been an important factor in the persistent heavy nature of the rain: a similar system in winter would not have produced so much rain. The heaviest rain fell later in the morning ( UTC) within the warm sector, but now it was aligned along the edge of the upper-level trough which was forcing ascent. The rain cleared from southeast England around UTC (Figure of Prior and Beswick,, this issue, p. ) and, by the afternoon ( UTC), most of the rain was falling within the air ascending above hpa into the cloud hook. The RDFD trajectory analysis in Figure (b) shows that the air that had experienced most rapid ascent over the preceding hours was over Gloucestershire at UTC. A vertical section across the front (Figure (c)) shows the depth of the WCB in this case. Back-trajectories from the X within the WCB (Figure (d)) originated from the Atlantic north-west of Spain four days previously, but remained at low levels until Figure. A series of -hour forecasts for Case C on July, from the Met Office operational North Atlantic and European model. Details as in Figure.

8 (c) (a)... PV (PVU) T=-. days hpa Release on UT // Wet bulb PT (C) T=-. days UT // reaching northern France, when the lowlevel cyclone and ascent developed rapidly. Conclusions concerning the origin of the three rainfall events In each of the three major rainfall events which led to flooding during summer, rainfall totals were extremely high because the heavy rain was persistent, although short-term rainfall rates were not generally as high as observed in severe thunderstorms T=-. days hpa Release on UT // Mesoscale model forecasts indicate that the pattern of rainfall was determined by the development of warm conveyor belts associated with mid-latitude cyclones. Isolated cells of deep convection did not play a major role, although convection was active within the warm sectors associated with potential convective instability above the warmest air at low levels. Back-trajectories calculated from analyses draw out the same structures as seen in the mesoscale forecasts. They also show that air within the WCBs had similar properties in all three cases (wet bulb potential E.N p=hpa T=-. days UT // (.N,.E) Distance (km) (.N,-.E) Longitude (degrees) WCB Figure. Reverse domain-filling trajectory analysis for Case C, showing properties of air at UTC on July. As in Figure except (a) potential vorticity carried along -day trajectories arriving on the hpa level; (b) change in pressure for -day trajectories arriving on the hpa level; (c) wet bulb potential temperature carried along -day trajectories; (d) -day back-trajectories from hpa over Gloucestershire, shown by an X in panels (a) (c). Figure. Thermodynamic sounding, plotted as a skew-t diagram, and wind profile from Herstmonceux, south-east England, at UTC on July. In addition to isobars and isotherms, dry adiabats (green), moist adiabats (blue) and isolines of specific humidity (maroon) are shown. Weak easterly boundary layer flow was overlain by southerly flow of a deep near-saturated moistadiabatic layer. Column precipitable water content was. mm. (Courtesy, University of Wyoming.) (d) (b) temperature, θ w C) and clearly originated from low levels over the Azores region. It is important to note that Atlantic SSTs in the Azores region were anomalously low (Figure (c)), as was low-level temperature there, relative to the summer climatology. Consequently, it does not follow that the air was exceptionally laden with moisture because temperatures and therefore humidity saturation mixing ratios had increased at the air-mass origins. This is a mechanism frequently proposed when hypothesizing about possible increases in rainfall extremes with climate change, but does not apply to the summer cases. The rainfall was persistent because it was caused by ascent within mid-latitude cyclones, rather than isolated deep convection. The cyclonic systems themselves were not exceptional in their structure and were not explosive in their development. In all three cases, they were associated with an almost stationary upper-level trough and therefore propagated slowly north-eastwards. However, cyclonic systems of this nature are unusual in summer. The air drawn into the WCB was warmer and therefore carried higher column moisture content than would occur in other seasons, even though its θ w was not anomalously high for summer conditions near the Azores. In order to explain why summer was so wet in western Europe, it is necessary to explain why the upper-level trough stalled over the UK and Ireland for so much of the summer. The upper-level trough was part of a Rossby wave pattern around the Northern Hemisphere that remained remarkably persistent throughout two periods, June and July. The mechanism of Rossby wave propagation and sources of wave activity have been discussed. Such Rossby waves can exist freely without any forcing. It is plausible that the existing wave structure could resonate with weak forcing, producing a tendency to lock the phase of the wave pattern relative to the ground. This renders it very difficult to attribute the existence of the wave pattern in summer to any particular forcing. Certainly, idealized experiments with numerical models are required to explore candidate mechanisms and rule out those that do not elicit the appropriate response. Having said this, it then raises the question: why do large amplitude Rossby wave patterns not occur more often during summer if all the candidate processes are in place? Similar patterns appear more typical in other seasons. For example, Blackburn and Hoskins () and Hoskins () have related the high UK rainfall in autumn to a different stationary Rossby wave pattern. Severe weather events during November have been linked to the repeated occurrence of a global Rossby wave pattern (Shapiro and Thorpe, ). This is an exciting area of research with many open questions. At present it is not possible to predict whether Weather September, Vol., No.

9 Weather September, Vol., No. the frequency or amplitude of Rossby waves on synoptic to planetary scales will change as climate changes. Climate model simulations of storm-tracks have many deficiencies, not least that storms, such as the three studied here, would not be well resolved. Therefore, there is currently very low confidence in predicting whether extreme rainfall events of a similar nature to those occurring in summer will become more likely in the future or not. Simple thermodynamic arguments ignoring circulation changes are insufficient to describe such extreme events because the dominant process lies with anomalies in large-scale wave dynamics. Climate simulations using higher resolution global models are now beginning to be performed, to address precisely these issues. Acknowledgements We thank Brian Hoskins for useful discussions during the early stages of this study. ECMWF operational analyses were supplied via the British Atmospheric Data Centre (BADC). NCEP Reanalysis Derived data and Interpolated OLR data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their website at noaa.gov/. Nicholas Klingaman kindly provided the plots in Figure. References Blackburn M, Hoskins BJ.. The UK record-breaking wet autumn. UGAMP Newsletter, No., pp (March ). Available online at reading.ac.uk/~mike/extremes/ Ding Q, Wang B.. Circumglobal teleconnection in the Northern Hemisphere Summer. J. Climate :. Grahame NS, Davies P.. Forecasting the exceptional rainfall events of summer and communication of key messages to Met Office customers. Weather :. Hand WH, Fox NI, Collier CG.. A study of twentieth-century extreme rainfall events in the United Kingdom with implications for forecasting, Meteor. Apps. :. Hanna E, Mayes J, Beswick M, Prior J, Wood L.. An analysis of the extreme rainfall in Yorkshire, June, and its rarity. Weather. :. Hoskins BJ.. Atmospheric processes and observations. Phil. Trans. R. Soc. Lond. A :. Hoskins BJ, Ambrizzi T.. Rossby wave propagation on a realistic longitudinally varying flow. J. Atmos. Sci. :. Hoskins BJ, Karoly DJ.. The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci. : -. Methven J, Arnold SR, O Connor FM, Barjat H, Dewey K, Kent J, Brough N.. Estimating photochemically produced ozone throughout a domain using flight data and a Lagrangian model, J. Geophys. Res. :, doi:./ JD. Prior J, Beswick M.. The exceptional rainfall of July. Weather. :. Rodwell MJ, Hoskins BJ.. Monsoons and the dynamics of deserts. Q. J. R. Meteorol. Soc. :. Sardeshmukh PD, Hoskins BJ.. The generation of global rotational flow by steady idealized tropical divergence. J. Atmos. Sci. :. Shapiro MA, Thorpe AJ.. THORPEX International Science Plan. Available online at prog/arep/thorpex/ Wallace JM, Gutzler DS.. Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev. :. Weather Log, June. Weather (): i iv (August). Weather Log, July. Weather (): i iv (September). Correspondence to: Mike Blackburn, NCAS-Climate, Department of Meteorology, University of Reading, PO Box, Earley Gate, Reading, RG BB, UK m.blackburn@reading.ac.uk Royal Meteorological Society and Crown Copyright, DOI:./wea. The contribution of Nigel Roberts was written in the course of his employment at the Met Office, UK, and is published with the permission of the Controller of HMSO and the Queen s Printer for Scotland. Flooding of the River Severn to the north of Gloucester, July. ( John Clayton.)